Understanding U.S hospital billing practices - part 1
So unless you live here or aren't interested in healthcare (both equally valid excuses), the U.S. government just released Medicare costs and payments made by hospitals around the country. The New York Times subsequently published a number of articles about how hospital bills vary in the U.S. It was brought to my attention by two of my very dedicated followers, known affectionately as The Big R and Hruby the Fondo Destroyer. These particular articles are short and I think you should definitely give them a read. Briefly, they discuss that:
- Government data shows that hospitals charge wildly differing amounts for the same procedure
- The findings are likely to intensify a long debate over the methods that hospitals use to determine their charges. (I'm looking at you, Chargemaster).
- Experts say it is likely that the people who can afford it least are getting hit with extremely high hospitals bills that may bear little connection to the cost of treatment. Gee, low income earners getting the shaft? That's a first.
- Hospitals submitted bills to Medicare that were, on average, about three to five times what the agency typically pays to treat a condition, and variations between what hospitals charge may be even greater. Confused? You aren't alone.
I'll also add that TIME published an amazing article (sent to me by a rising star in the D.C. consulting biz) which goes into much greater detail about the many things that are broken with the U.S. medical system. It's quite long, but worth reading for anyone who cares to learn about the Medical industry is the largest spender in the U.S (much more than Energy or the Military-Industrial complex).
This leads us to a number of questions such as "how do hospitals determine their prices?", "why do they differ so widely?" and "what the f**k?" (2 of those questions are from the article above). Of course, to answer these questions intelligently would require a deeper understanding of the U.S. medical system, than I have. Moreover, after having conversations with professional practitioners, even they commented that (I'm slightly paraphrasing here) the system is arbitrary and generally speaking, a clusterf**k. So while I can't quite provide a scholarly dissertation on the matter, what I can do is analysizeTM. If only I had access to the data. Unfortunately, this type of stuff isn't always released to the public, not least because it contains potentially damaging information.
The data underlying this article is publicly available and can be found here: Medicare Provider Charge Data. Yes, I realize that the way I ended the previous paragraph did not lead you to believe that the data was in fact public. That's the power of words.
Let's take a quick look at this dataset. The data covers bills submitted from virtually every hospital in the country in 2011 for the 100 most common treatments and procedures performed. The two key variables in this data set are payments made by Medicare (to the hospital) and charges billed by hospital. We'll see that there is great variation in these charges across procedure, hospital and state.
With this much data (over 160,000 records) there are many ways to slice and dice it. Given that, I'm going to dedicate the next few posts to this data set. I hope to shed some light on the discrepancies between charges for various procedures in different hospitals around the country. So let's dive in.
For this first post, we'll begin with a big picture view. How do the amounts of charges by hospital and by Medicare vary, state to state? Since there are 100 different treatments in this data set, I'll choose two of the more ubiquitous ones - diabetes and heart related procedures.
Note that the comparison in the following maps is relative to each state. That is, the categories (arbitrarily, 0 through 10) are equally spaced out and therefore are only as extreme as the most extreme states. For example, if one state spends $100, and the other 47 states (sorry Hawaii and Alaska, not this time) spend $1, then the sample will be split into 2 - above $50 and below $50. If that $100 state was instead $10, then the sample would be split into above $5 and below $5, making the discrepancy less extreme. Make sense? If not, that's fine, just enjoy the pictures. This is common practice when doing a visual analysis, so remember to note that next time you look at color-coded visualizations.
Now the maps:
Interesting. It appears as though only California gets charged the most, and pays the most out (on average) for diabetes treatments. It looks like New York and Wyoming (of all places!) charges a lot, but gets little covered by Medicare. Maryland especially has a large discrepancy in this balance. Conversely, both New Jersey and Nevada get more Medicare coverage than their hospitals charge.
I wonder if that is consistent with people's expectations about a state's demographics, health composition, level of hospital expertise, technological advancement etc... It's curious that New York and Wyoming are similar in this regard, given how vastly different their underlying populations are. Perhaps this goes to show you that the system is completely arbitrary.
Let's take a look at heart related treatments.
Heart to tell a difference (I couldn't help myself) from the diabetes map above, but slight nuances exist. For example, Wyoming and Nevada are less extreme in their costs for Heart treatments. California, New York and New Jersey still lead the way in their respective expense categories. In general the Heart map appears to be more balanced than the Diabetes map. I will note that procedures that fall under the category "HEART" are much more numerous than "DIABETES" and therefore a lot of variability will be smoothed out since the sample is larger (law of large numbers at work). It might make more sense to visualize individual treatments, rather than groups, in order to detect outliers.
However, given the length and density of this post, I'll save that analysis, including one at the state level, for another post.
Doctors, practitioners and other healthcare enthusiasts - don't lend me your ears, but instead let me know what you'd like me to analyze. There is a TON of information here and I wouldn't know what the most insightful thing is. Let me know, either in the comments, via facebook, email or the ever trusty raven mail.
That's it for now!